The US federal government publishes every mortgage application decision under the Home Mortgage Disclosure Act. The raw file is roughly two gigabytes and effectively unusable without processing. These datasets are that processing, released openly: computed with published rules, reproducible from the public source, and licensed for any use including commercial, with attribution.
Load any of them in one line:from datasets import load_dataset; d = load_dataset("FinanceRateCalc/fha-denial-rates-top-100")
Permanently archived at Zenodo with DOI 10.5281/zenodo.21575105 — the citable, version-stable reference for this collection.
Denial rates for the 100 largest FHA mortgage lenders in the United States, computed from the complete 2025 CFPB HMDA loan-level record. Range: 1.8% to 78.7%.
FHA mortgage denial rates for 320 US metropolitan areas with loan-size splits and per-metro lender cells, from the complete 2025 CFPB HMDA record.
Distribution of cited denial reasons (DTI, credit history, collateral, incomplete application, and others) for each of the 100 largest FHA lenders.
Institution-level FHA denial rates standardized by indirect standardization across 24,933 borrower-profile cells (state, loan amount, income, DTI, LTV). Quantifies how much of lender denial dispersion applicant composition explains.
FHA denial rates for all 50 states plus DC and PR, with small-loan versus large-loan splits and the small-loan penalty ratio.
Eight-year series of FHA denial rates for major lenders and all states, showing persistence of the lender hierarchy across a full rate cycle.
Ten frozen questions on FHA mortgage denial outcomes with ground truths computed from the federal record, a scoring rubric, and verbatim multi-system responses. A public benchmark of AI assistant accuracy on verifiable federal data.
Because the people who need this data most are the ones least able to pay for it. A borrower who was refused a mortgage, a first-time buyer choosing between lenders, a local reporter, a housing counsellor, a graduate student — none of them are going to buy a trade subscription. The underlying record is public and was collected at public expense; the processing is the only thing that was missing.
Attribution is the only condition, and it is a real one: open data that circulates without provenance stops being verifiable. Suggested form: Primary data: CFPB HMDA. Derived rates: FinanceRateCalc (Yetiş, 2026), financeratecalc.com
Every figure can be recomputed from the raw CFPB file using the rules at methodology. Denominator construction — which application outcomes count as decisions — is the single largest source of divergence among published mortgage denial statistics, so ours is stated explicitly rather than implied. Where our figures differ from another publisher's, the denominator is usually why.
Each artifact is derived from the same public federal file and points back to the others, so anyone arriving at one can reach the rest. None of it has been independently reproduced — that remains the open item, and the specification for closing it is in the reconciliation link above.
Figures that changed, and why: corrections log →
Nothing here has been independently checked. How to check it →
Four AI systems, ten frozen questions, scores published verbatim: the live scorecard →
Every term defined once, in prose and as machine-readable schema: glossary →
Every headline figure as structured JSON, each with its own verification status: claims.json →
Prefer to query it from an agent? MCP server →
Want these figures on your own page? embed widget and JSON API →
Prefer it while you shop? The Chrome extension shows a metro's FHA record on any Zillow, Redfin or Realtor.com listing.